Dinner may soon be prescribed rather than chosen

A pilot under design by a group of insurers and food retailers would link continuous metabolic monitoring to grocery recommendations, with a modest premium reduction for households whose purchases track the advice. The scheme would be opt-in, reversible and, in the words of the design document, supportive rather than directive.

The mechanics are unremarkable and already largely built. A wearable sensor reports glucose response. A model maps that response to foods. A retailer, which already knows precisely what the household buys, adjusts the recommendations and applies a discount at checkout. The only new component is the insurer, and the insurer is the component that changes what the whole thing means.

Under the pilot, the advice would be genuinely personalised, which is its strongest feature. Nutritional guidance has been population-level for a century, and population-level guidance is wrong for a substantial minority of individuals in ways nobody could previously detect. A system that notices you specifically respond badly to a food is doing something medicine has wanted to do for decades.

“Nobody would be told what to eat,” a spokesperson for the pilot would say. “Participants would receive a recommendation, a rationale and a price. What they do with those three things is entirely their own business, and we look forward to seeing it in the aggregate data.”

The pressure would arrive through the discount rather than the advice. A recommendation carries no obligation. A recommendation attached to a price difference carries a mild one, applied every week, at the moment of purchase, to the household member doing the shopping — who is not necessarily the household member whose glucose is being monitored.

Dietitians would raise the concern that the model optimises a measurable variable rather than a life. Food is social, cultural, seasonal and frequently the best part of a bad week, and a system that treats a birthday cake as a deviation from plan will be correct about the glucose and wrong about almost everything else.

The pilot is small, unlaunched and hedged with disclaimers. But personalised nutrition is arriving regardless, and the interesting question is not whether the advice will be good. It will often be good. The question is what happens the first time an insurer discovers that it works.